EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng. Struct. Dyn. 27, 845867 (1998)

MODELLING ASPECTS OF STRUCTURES ISOLATED WITH THE
FRICTIONAL PENDULUM SYSTEM
JOSE L. ALMAZA N , JUAN C. DE LA LLERA* AND JOSE A. INAUDI`
m
Department of Structural Engineering, The Catholic University of Chile, Casilla 306, Correo 22, Santiago, Chile
` Department of Civil Engineering, University of California at Berkeley, Berkeley, CA, U.S.A.
SUMMARY
Dierent modelling aspects of structures isolated using the frictional pendulum system and subjected to earthquake
ground motions are studied herein. Although the vertical dynamics of these structures is given special emphasis, other
eects such as large isolator deformations and bidirectional input motion are also considered. Dierent structural models
of the FPS are developed and tested for single-storey structures and a real four-storey building frame; among them, an
exact formulation of the FPS forcedeformation constitutive relationship is presented. Results show that global building
responses can be computed within 20 per cent error in the mean using a simplied model that ignores the vertical motion
of the building; however, structural member deformations and forces need to be computed using a model that considers
such motion. This is of particular importance when there exist correlation between the horizontal and vertical
components of ground motion. Further, a physical model of the FPS is introduced and used to determine the response of
a real four-storey frame, including uplift and downward impact. Results from this analysis show that local column
responses may vary substantially depending on the stiness of the isolation storey and the presence of a mass at the
isolation level. Such mass is capable of ltering the large increase in column shear that results from the impact of the
structure after uplift. Uplift occurs at several instants of the response of the structure considered, leading to an increase in
column base shear as large as 3 times the shear obtained by ignoring the vertical dynamics of the building. 1998 John
Wiley & Sons, Ltd.
KEY WORDS: frictional pendulum system; base isolation; physical model; uplift; normal forces
INTRODUCTION
Coulomb friction has proven to be a simple and reliable mean of vibration isolation and energy dissipa-
tion in earthquake engineering applications. Frictional forces developed at the interphase of two
sliding materials are well understood in their functional dependency with contact pressure and sliding
velocity.`
Among the seismic isolation devices based on friction as means of energy dissipation, a remarkable kind is
the widely known Frictional Pendulum System (FPS).`` This system consists of a spheric stainless steel
surface and a lentil-shaped articulated slider covered by a Teon-based high bearing capacity composite
* Correspondence to: Juan C. de la Llera, Department of Structural and Geotechnical Engineering, Ponticia Universidad Catolica de
Chile, Casilla 306, Correo 22, Santiago, Chile
Ph.D. Student
Assistant Professor
m
Research Associate
Contract/grant sponsor: FONDECYT; Contract/grant numbers: 1950987, 1971078
Contract/grant sponsor: FONDEF; Contract/grant number: D96I1008
CCC 00988847/98/08084523$17)50
1998 John Wiley & Sons, Ltd. Received 1 August 1997
Figure 1. Schematic view of the FPS and equilibrium diagrams in the 2-D and 3-D cases: (a) schematic section of a typical FPS;
(b) planar model; (c) three-dimensional model
material (Figure 1(a)). Perhaps, one of the most relevant features of the FPS is that residual displacements in
the isolation are reduced due to the self-centering action induced by the concave spherical surface. Residual
displacements are an important drawback of other sliders in general. A complete description of the
advantages of the FPS system relative to other isolators may be found elsewhere.`"
Several structures have been built in recent years using FPS isolators or devices alike. An interesting, well
documented, and recent case is the historic Ninth Circuit U.S. Court of Appeals building located in San
Francisco,` damaged during the Loma Prieta earthquake in 1989 and retrotted recently using 256 FPS
isolators.
Over the past 10 years the FPS has been studied analytically and experimentally by a number of
researchers.` " ' ` Earlier studies developed simplied analytical models capable of representing the pre-
dominantly bilinear behaviour of these isolators.` " They also showed that these models led to results in
agreement with measured global responses of shake table tests. Recently, large-scaled tests have been
performed at NCEER, Bualo, on a six- and seven-storey steel frame supported by FPS.` Again, results of
these tests have been correlated well with global analytical results. Further, some experimental results`
insinuate that torsional eects would be naturally controlled in structures with FPS, at least for mass
eccentric systems.
In spite of the thorough work already done to understand the dynamic behaviour of structures isolated
with the FPS system, there are still important aspects of the behaviour of such structures that need further
846 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
investigation. Indeed, the vertical vibration of a structure supported by FPS is especially important, since the
local behaviour of the isolators is dependent on the contact pressure between the slider and the steel surface.
Although such eect have been shown to aect only slightly the global response of a 2D-frame,` they do
aect the local behaviour and design of the isolators and, hence, deserve special attention from the designer.
Moreover, variability in the contact pressure among isolators due to overturning or vertical input motion
may induce coupled lateral-torsional vibrations and uplift that need to be better understood. Besides, given
the powerful computational tools available today it is possible to formulate an exact solution of the motion
of system isolated with the FPS. Such solution can be built in consecutive steps of increasing complexity to
better appreciate the approximations implicit in the models available in practice today for predicting the
response of FPS isolated structures.
Consequently, the objective of this investigation is to develop a theoretically exact model of the FPS
system and relate such model with those developed earlier and used in practice today. Among the modelling
issues studied herein, the vertical dynamic behaviour of structures with the FPS system is given special
attention. Other modelling aspects, such as biaxial eects, large deformations, and uplift, are also dealt with
and analysed. Besides, a physical model of the FPS is presented and briey discussed; such model provides
a simple way to account for uplift and coupled lateralvertical vibrations of structures isolated with the FPS.
Finally, a multistorey building example is analysed and the results compared among dierent models. This
example points out some important issues of the behaviour of these systems that need to be considered by the
designer.
Structural responses are studied for a group of six selected earthquake motions. Among them, three recent
California earthquake records are considered, which show correlation between lateral and vertical compo-
nents. This correlation turns out to be an important parameter in assessing the signicance of lateralvertical
coupling in the response of the structure. Because such correlation is not as much present in other
conventional motions used in earlier investigations of the FPS, it is interesting to see the implications of such
eect in the global as well as local response of the structural system. Furthermore, the large displacements
observed in these near-eld earthquake records may induce large displacement eects on the FPS that need
to be considered appropriately in the modelling.
FORCEDEFORMATION CONSTITUTIVE RELATION OF THE FPS
An important aspect in formulating the complete equations of motion of systems isolated with the FPS,
is to state the exact forcedeformation constitutive relationship of the isolator. Indeed, the dierent
levels of complexity involved in formulating this relationship are directly related with the accuracy of
the model. In the formulation presented displacements are imposed to the FPS and forces are retrieved
from it.
Let us state rst the well-known forcedeformation relationship of the FPS system in one dimension and
small deformation (Figure 1(b)), which results from horizontal equilibrium of the isolator.` "
f"

x
'
#jsgn (xR
'
) (1)
where represents the weight of the structure acting on the isolator; x
'
and xR
'
are the lateral deformation
and velocity of the isolator relative to the ground, respectively; R

is the radius of the spherical surface; j is

the friction coecient; sgn(xR
'
) is the signum function, i.e., equal to !1 or #1 depending if the velocity is
negative or positive, respectively; and f is the horizontal restoring force of the isolator. Equation (1) is derived
assuming that the angle 0 between the normal and the vertical direction is small (Figure 1(b)) thus,
cos(0)"1 and j tan(0)+0. Based on this assumption the normal contact force N, is implicitly assumed
equal to . Naturally, such approximation is valid only when the vertical component of ground motion is
ignored.
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 847
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Equation (1) may be extended to the case of bidirectional motion of the isolator by stating the conditions of
dynamic equilibrium in the horizontal plane, leading to:
f"

q
"
#j
q
"
] q
"
]
(2)
again, f"[ f
V
f
W
]' is the horizontal restoring force of the isolator, where f
V
and f
W
are the cartesian
components of this force in the x and y directions, respectively; q
"
"[x
'
y
'
]' is the horizontal deformation
of the isolator relative to the ground with x- and y-components x
'
and y
'
; and q
"
"[xR
'
yR
'
]' is the horizontal
velocity of the FPS. The similarity between equations (1) and (2) is obvious when identifying sgn (xR
'
) with
q
"
/ ] q
"
], the latter corresponding to a unitary vector tangent to the horizontal isolator trajectory. Both
equations (1) and (2) are limited to small isolator deformations.
By using the actual displaced geometry, equation (2) may be extended to the most general case of
large isolator deformations. Horizontal dynamic equilibrium of the isolator in that case leads to
(Figure 1(c)):
f"
N
j
q
"
#jN
q
"
] q ]
(3)
where N is the magnitude of the normal contact force, j is the radius of curvature of the concave surfacethis
radius is constant and equal to R

for the FPS system; however, the formulation presented is not limited to
such case; and q "[xR
'
yR
'
zR
'
]' is the true velocity of the isolator relative to the ground, where zR
'
is its vertical
component. This latter velocity term results from imposing the kinematic constraint corresponding to the
motion of the isolator on the spherical (or arbitrary) sliding surface; its value depends on the horizontal
displacement q
"
and velocity q
"
of the FPS.
It is important to note in equation (3) that the reaction N is no longer constant and equal to the weight .
The value of this reaction is continuously varying during the motion of the system due to the vertical
component of ground motion and the lateralvertical coupling that exists between the horizontal and
vertical motion of the FPS. Moreover, since the isolator deformation q
"
and velocity q
"
are imposed by the
structure, the only unknown at the element level is the magnitude of the normal reaction N.
In general, equations (1)(3) show that the horizontal restoring force of the isolator comes from
the superposition of two dierent actions, one resulting from the pendular eect associated to the
centering action of the weight , f

, and the other, resulting from the frictional forces developed at the
isolator interface between the slider and surface, f
I
. Thus, equation (1)(3) may be expressed more
synthetically as
f"f

#f
I
(4)
Equations (1)(4) represent the horizontal restoring force of the isolator when it is sliding on the concave
surface. During the sliding phase of motion, the frictional force f
I
always opposes to the direction of
the velocity implying that the total reaction R at the sliding surface, whose component of the tangent plane
is the frictional force, is always located on the surface of the friction cone (Figures 1(b) and 1(c)). This reac-
tion changes direction continuously following the isolator trajectory; however, since the angle between
the normal to the surface nL and the reaction R at a given point is xed "tan (j), as soon as
the angle between the resultant external force and the normal nL is larger than , the magnitude of the iso-
lator velocity will increase. On the other hand, if the angle between the resultant of the external forces
and the normal nL is smaller than , the isolator velocity will decrease until it stops and sticks to the concave
surface. From there on, the structure will function as a xed-base system. In the sticking range, the reaction
R will be that required to maintain equilibrium with the external resultant. During the actual motion of the
system, the isolator will shift between these two phases, i.e., a sticking phase (phase I) and a sliding phase
(phase II).
848 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
STRUCTURAL SYSTEMS AND RESPONSES CONSIDERED
The structural model considered rst corresponds to a single-storey building isolated with one FPS
(Figure 2(a)). The masses of the structure and the base are m
'
and m
'
, respectively, and their ratio will be be
dened as :"m
'
/m
'
. As shown in the gure, the columns between the oor and isolation levels are modelled
by springs and dashpot elements in three perpendicular directions, the two horizontal x and y, and the
vertical z.
Figure 2. Schematic view of the systems considered: (a) single-storey system considered in parametric study; and (b) four-storey building
frame used in example
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 849
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Table I. Structural models considered in this study
Structural model
Case Constitutive
model relationship Simplied Exact Observations
I Equation (2) ( N"
II Equation (2) ( N" (1#u
EX
/g)
III Equation (3) ( Exact N
IV Equation (3) ( Exact N
Two main structural models are considered in this study. First, a simplied structural model which
neglects the vertical response of the system by imposing innite axial stiness in the building columns. And
second, an exact model which considers the lateral and vertical motion of the system using the actual
horizontal as well as vertical stiness of columns. In turn, both structural models are combined with dierent
models of the forcedeformation constitutive relationship of the FPS system, leading to the four analysis
cases described in Table I.
Thus, case I represents a building with innite axial stiness in columns and connected to an FPS isolator
modelled assuming small deformations and a constant normal contact force equal to the weight above the
isolator, "(m
'
#m
'
) g. Case II is similar to I but includes the vertical component of ground acceleration
in the computation of the normal contact force. This model has been considered in earlier investigations.`
Furthermore, case III is identical to cases I and II but considers the exact value of the normal contact force
N assuming large deformations in the isolator but innite vertical stiness of columns. Finally, case IV
considers the exact solution including large deformations of the isolator and nite vertical stiness of the
structure.
The deformations of the structure relative to the ground will be denoted as r"[x
'
y
'
z
'
]'"[r
"
z
'
]', and
the deformations of the base relative to the ground as q"[x
'
y
'
z
'
]'"[q
"
z
'
]', where r
"
and q
"
represent
a partition containing the two horizontal degrees of freedom of the structure and base, respectively. The
corresponding mass, damping, and stiness matrices of the xed-base structure will be denoted as M
'
, C
'
and
K
'
; a consistent partition of these matrices with r
"
and z
'
is as follows:
M
'
"

M
'"
0
0
m
X

, C
'
"

C
'"
0
0
c
X

, K
'
"

K
'"
0
0
k
X

(5)
where M
'"
, C
'"
, and K
'"
are the mass, damping, and stiness matrices associated with the horizontal degrees
of freedom of the superstructure; m
X
"m
'
, the translational mass of the structure, c
X
"[c
'
, the damping
coecient of the structure; and k
X
"[` k
'
, the lateral stiness of the structure, where ["c
`
/c

represents
the ratio between the vertical and horizontal vibration frequency of the structuretypical values for [ in
frame buildings range between 5 and 15.
System responses were computed for six dierent free eld ground motion records. The motions con-
sidered were Vin` a de Mar and Llolleo (Chile, 1985), Corralitos (Loma Prieta, 1989), Sylmar and Newhall
(Northridge, 1994), and Kobe University (Kobe, 1995). The three ground acceleration components of these
motions are presented in Figure 3. Since the selected recorded motions of the Kobe earthquake correspond
to ground velocity, ground acceleration have been computed by numerical dierentiation of this original
record. The reader is warned that in the case of the vertical component of the Llolleo record (Chile, 1985), the
initial part of the vertical component has been omitted due to the apparent faulty sensor functioning; in spite
of that, the remaining unperturbed section of the record has been included because of its great interest for
structures subjected to earthquakes originated in subduction zones and founded on sti soils.
850 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 3. Ground motion records considered in this study
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 851
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
The results presented in this study consider a constant value of the friction coecient`` j"0)07.
Although including the variability of j with velocity and pressure as proposed in earlier work does not
imply any extra diculty in the model, it seemed unnecessary due to the minor eect that the variation in
j has on the peak response values of the system. Further, a constant damping ratio "0)05 was considered
for the superstructure in all analysis cases.
The system responses considered in this study are: (1) the normalized horizontal deformation q
"
"q
"
/R

of
the isolation level relative to the ground, (2) the normalized horizontal restoring force f"f/W of the isolator,
(3) the normalized base shear V "V/(m
'
g) of the structure, and (4) the normalized contact reaction in the
isolator N "N/W normal to the sliding surface. Because of the mass m
'
lumped at the isolation level, it is
important to emphasize that the horizontal restoring force on the isolator f is dierent from the total base
shear V of the structure; the dierence being the inertia force generated at the base.
EQUATIONS OF MOTION
Next, the general equations of motion of the two structural models described above are stated and
interpreted. For each structural model, the sticking (phase I) and sliding (phase II) phases of the motion are
identied and the corresponding equations presented. In the most general case, the degrees of freedom of the
system are ve; three degrees of freedom of the superstructure and two degrees of freedom of the isolated
base. The degrees of freedom of the structure and base will be denoted as d"[r
"
z
'
q
"
]'"[r q
"
]', and the
corresponding velocities and accelerations as d and d .
During the sticking phase, the governing equations of motions of the system correspond to that of
a xed-base structure. For a structure with innite vertical stiness in columns, the equations are:
M
'"
r
"
#C
'"
r
"
#K
'"
r
"
"!M
'"
u
"
!K
''
(1 : 2, 1 : 2) q
"
(6)
and for a structure with nite vertical stiness in columns,
M
'
r#C
'
r #K
'
r"!M
'
u

!K
''
q (7)
where K
''
represents the o-diagonal matrix coupling r and q; and the notation (i : j, k : l) represents
the submatrix delimited between rows i and j and columns k and l. These equations are essentially
identical to those of a xed-base structure but include the shift in position of the slider q at the time of the last
sticking.
Pre-multiplying equations (6) and (7) by the inverse of the mass matrix M
'
, equation (6) can be written in
parametric form as
r
"
#2c

r
"
#c`

r
"
"!u
"
!c`

q
"
(8)
and equation (7) as

r
"
z
'

#2c

I
`
;
`
0
0 [
r
"
zR
'

#c`

I
`
;
`
0
0 [`
r
"
z
'

"!

u
"
u

!c`

I
`
;
`
0
0 [`
q
"
z
'

(9)
In equation (9), the vertical displacement of the isolation base z
'
is such that it satises the kinematic
constraint imposed by the sliding surface, i.e., G(x
'
, y
'
, z
'
)"0. For convenience, z
'
is chosen as the
dependent variable of q
"
, implying that z
'
"g (q
"
).
852 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
On the other hand, the FPS horizontal restoring force f can be directly computed from equilibrium at the
isolation base for both structure models as
f"![K'
''
(1 : 2, 1 : 2) r
"
#K
'
(1 : 2, 1 : 2) q
"
#C'
''
(1 : 2, 1 : 2) r
"
#M
'"
u
"
] (10)
where K
'
represents the submatrix of stiness associated with the deformations q of the isolated base; C
''
is
the damping submatrix coupling the r and q velocities; and M
'"
is the mass submatrix associated with q
"
.
Equations (5)(9) can be used as long as the horizontal restoring force f of the FPS does not exceed the
frictional resistance developed by the isolator. The condition of sticking can be simply stated saying that the
reaction at a given point on the surface must remain within the friction cone (Figure 1(c)). This leads to the
following condition for sticking:
] R!N n ](jN (11)
where R is the total reaction developed at the sliding surface of the isolator; N"R ) n is the magnitude of the
reaction normal to the tangent plane at the point considered; and ] ( ) ) ] represents the euclidean norm of the
vector. For instance, the value of the vertical component of the reaction, R
X
, is obtained from equilibrium
with the vertical action imposed by the structure onto the FPS f
X
"k
X
(z
'
!z
'
)#c
X
zR
'
, the vertical inertia
force acting on the base m
'
u
X
, and the weight of the structure (m
'
#m
'
) g. In the case of small deformations,
equation (11) is simplied to

f!

q
"

(j (12)
Next, the complete equations of the sliding motion of the structural system shown in Figure 2(a) are stated.
The degrees of freedom considered are 4, both horizontal deformations of the structure and base, or 5, i.e.,
d"[r q
"
]', if the vertical deformation of the structure is included. Recall that the vertical deformation,
velocity, and acceleration of the base relative to the ground z
'
, zR
'
, and z
'
, are kinematically dependent of the
horizontal deformations q
"
of the FPS.
In order to reduce the number of equations presented, a general formulation of a structure sliding on the
FPS is presented; the formulation for the simplied structural system having innite axial stiness in the
columns can be obtained directly from the more general equation presented by eliminating the row and
column corresponding to the vertical degree of freedom z
'
of the structure:

M
'
0
0
M
'"

r
q
"

C
'
C'
''
(:, 1 : 2 )
C
''
(:, 1 : 2)
C
'
(1 : 2, 1 : 2)
r
q
"

K
'
K'
''
(1 : 2, 1 : 2)
K
''
(:, 1 : 2)
K
'
(1 : 2, 1 : 2)
r
q
"

"!

M
'
0
0
M
'"

B
"
u

K
''
(:, 3)
0
C
''
(:, 3)
0
z
'
zR
'

0
f
(13)
where B
"
represents the inuence matrix of the input, C
'
is the damping matrix associated with the base
velocity q ; and f is the horizontal restoring force of the isolator computed using equations (2) or (3) depending
on the constitutive model considered. The right-hand-side term including z
'
and zR
'
in this equation
represents the forces applied to the r degrees of freedom of the structure due to the constrained vertical
displacement and velocity of the FPS. As mentioned before, z
'
is computed directly from the kinematic
constraint z
'
"g (q
"
); zR
'
can also be computed directly from this equation:
zR
'
"gR (q
"
, q
"
)"

cg
cq
"

'
q
"
"
cg
cx
'
xR
'
#
cg
cy
'
yR
'
(14)
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 853
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Therefore, the r.h.s. term in equation (13) depending on z
'
and zR
'
, is actually dependent on the horizontal
displacements and velocities of the FPS. It is through this term and the horizontal restoring force f that
lateralvertical coupling is developed in the structural system. The parametric representation of equation (13)
is omitted here for the sake of brevity.`
COMPUTATION OF THE NORMAL REACTION N
The integration of the equation of motion (13) diers from that of a xed-base structure due to the
computation of the non-linear restoring force f of the FPS isolator (Equations (1)(3)). This force depends, in
turn, on the normal reaction N developed at the spherical sliding surface, which magnitude is computed from
dynamic equilibrium of the FPS in the direction of the normal n to the concave surface:
N"n
X
!A ) n #m
'
u

) n #m
'
a
L
(15)
where n
X
is the vertical component of the normal n to the surface, a
L
"q ) n is the normal acceleration of the
isolator relative to the ground;` and A"!m
'
(u

#r) is the dynamic interaction force between the

superstructure and base. Substituting A in equation (15) and dening the relative acceleration of the structure
with respect to the isolation r
''
"r!q , this equation leads to:
N"n
X
#(m
'
#m
'
) u

) n #(m
'
#m
'
) a
L
#m
'
r
''
) n (16)
Each term in equation (16) can be physically interpreted. For instance, the rst three terms correspond to the
normal force produced as if the system were a rigid body. The rst term is the component of the weight along
the normal direction; the second term is the component of the inertia force generated on the structure as
a rigid body by the ground motion; and the third reects the eect due to the spherical motion of the FPS
along the spherical isolator surface. The last term on the r.h.s. of this equation reects the eect that the
exibility of the structure has on the value of the normal reaction.
So far, the formulation presented assumes that no uplift is produced in the system during the motion,
i.e., N'0. Uplift will be introduced in a natural form later through a proposed physical model of the
FPS.
INTEGRATION OF THE EQUATIONS OF MOTION
During phase I of the motion of the structure, equations (8) and (9) represent the well-known linear motion of
a xed-base structure. For numerical simplicity these equations are integrated by rst transforming them
into rst-order dierential equations (state-space formulation) and using a rst-order hold in the integration.
Phase changes between sticking and slip are evaluated using equations (11) and (12); thus, for instance,
sticking of the structure is produced when zero velocity is reached at the FPS and the frictional force required
for equilibrium is less than the maximum frictional force. Next, the integration algorithm is briey described
for the exact model of the structural system considered; other structural model cases are analysed using the
same integration procedure.
Let us start assuming that the state of the system, i.e., the displacements d(t
I
)"[r (t
I
) q
"
(t
I
)] and velocities
d

(t
I
)"[r (t
I
) q
"
(t
I
)], are known at instant t
I
"k, where is the integration step. The integration of
equation (13) continues as follows: (1) determine the unitary normal vector n to the surface at x(t
I
); (2)
compute the normal acceleration a
L
;` (3) determine the interaction force A"K'
''
r#K
'
q#C'
''
r #C
'
q ; (4)
evaluate the magnitude of the normal reaction N using equation (15); (5) determine the horizontal restoring
force f developed by the FPS using equations (1), (2), or (3); (6) compute the vertical displacement and velocity
of the isolator, z
'
and zR
'
, by equation (14); (7) compute r and q
"
using equation (13); and (8) compute the state
in the next step, d (t
I>
) and d (t
I>
). The last step in the integration procedure depends on the time
854 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
integration method used. In this investigation a fourth-order RungeKutta method has been selected. In
order to use the method, equation (13) was transformed into a rst-order dierential equation.
ANALYSIS OF RESULTS
The objective of this section is to study parametrically the earthquake behaviour of the single-storey system
considered using dierent structural models. The results presented are intended to evaluate the eect of
specic phenomena present in the behaviour of structures isolated using the FPS, such as the eect of large
deformations, vertical input, vertical exibility of the structure, and bidirectional input motion. Although due
to the non-linear nature of the coupled equations of motion of the system, it is impossible to isolate and
evaluate independently the eect of dierent modeling assumptions, an eort has been made in presenting
results that show how dierent eects control the earthquake response of these systems.
Let us start considering the unidirectional lateral impulse response of the system with rigid and de-
formable superstructure. Shown in Figure 4 is a comparison between the normalized horizontal base
deformation xL
'
, the normalized base velocity xR
'
(t) /xR
'
(t"0), the normalized normal force NK, and the
normalized force deformation relationship of the isolator. The parameters of the rigid superstructure model
Figure 4. Comparison between the impulse response of the system including (

"0)5 sec, ["7, :"0)2) and neglecting (

"0 sec,
["R, :"0)2) the exibility of the superstructure
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 855
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
are

"0 sec., i.e., ["R, and mass ratio :"0.2; the corresponding parameters of the exible superstruc-
ture are

"0)5 sec., ["7, and :"0)2. These results show that dierences between the isolator deforma-
tions computed from both models are very small; however, slightly larger discrepancies, say 10 per cent, are
observed in the normal force as a result of the lateral and vertical exibility of the system. The change in
normal force relative to 1 in the latter model is due to the large displacement eects of the slider moving along
the concave surface, the normal acceleration thus produced, and the exibility of the system (equation (16)). It
is also responsible for the interesting waviness shown in the normalized forcedeformation relationship of the
exible system.
Because the normal contact force NK is intuitively expected to aect the earthquake response of the
structure, the dynamic variation of this force is evaluated in greater detail. Compared in Figure 5 are the
earthquake responses of the structure (

"0)5 sec, :"0)2, ["7) obtained from model cases I and IV. Both
models are subjected to the Newhall ground motion record; however, in case model I only the two horizontal
components are included as opposed to the three components in case model IV. The upper two plots show
the interaction between the two components of the frictional force f
I
in the isolator; the plots on the left of the
gure correspond to results obtained from the approximate model (case I) and those on the right to results
Figure 5. Earthquake response of case models I and IV subjected to the Newhall ground motion (

"0)5 sec, ["7, :"0)2) for: (a)

case model I; and (b) case model IV
856 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
obtained from the exact model (case IV). As it should be, the frictional forces in the simplied model describe
a circular interaction; however, since the normal force is continuously varying during the motion of the
structure in case model IV, sliding occurs inside as well as outside the circular interaction curve. Indeed, the
discrepancies observed between these two plots is a measure of the eect of the variable normal force NK.
Notice also that the friction forces during sticking (circles) maintain certain similarity in both case models as
opposed to the large discrepancies observed in these forces during sliding (phase II). The second row of plots
describes the normalized pendular component f

of the FPS horizontal restoring force versus the normalized

isolator deformation q
"
. The relationship is linear for the approximate model with NK"1, and diers from it
in the exact model due to the continuous variation in the normal contact force. The last row of plots shows
the normalized constitutive forcedeformation relating the x- and y-direction isolator force and deformation.
The waviness present on the two plots on the right results from the variation in the contact normal force
caused by the true modelling of the vertical dynamics of the structure. As shown later, the occurrence of these
spikes at large values of displacement can be explained by the correlation between peaks of the horizontal
and vertical ground motion components.
Mean spectra of maximum and minimum normalized NK are presented in Figure 6 for the six ground
motions selected. These spectra were computed using the exact structural model (cases IV) and the three
components of each ground motion. They are plotted as a function of the xed-base period of the structure
for values of vertical to lateral stiness ratio ["5 and 7, and three values of the mass ratio :"0)1, 0)2, and
0)5. As shown in the gure, the mean maximum values of NK range approximately between 1)6 and 2, implying
that amplications of 50100 per cent should be expected in the normal contact force during an earthquake.
On the other hand, minimum mean values of NK vary from about 0)2 to 0)6, implying that decreases of the
normal contact force of the order of 1/2 to 1/5 should also be expected. The peak values of NK tend to be
larger for the maximum and smaller for the minimum in the mid-period range, respectively. Further, these
extreme values are quite stable for dierent values of the vertical-to-lateral stiness ratio [ and mass ratio :.
Now that the variations produced in the value of the normal contact force NK have been evaluated, it is
important to study the eect that these variations may have in the earthquake response of the system. Let us
rst consider the case of a single structure subjected to the Newhall ground motion (Figure 7). The
comparisons presented here are again between the structural case models I and IV (Table I). The structural
system considered has a xed-base period

"0)5 sec, vertical to lateral stiness ratio ["7, and mass ratio
:"0)2. The results presented in the gure are the normalized x-direction isolation displacement xL
'
and base
shear K
V
, the magnitude of the normalized normal force NK, and the normalized x-direction forcedeformation
loop. Notice that the horizontal deformations at the isolation level are similar for both structural models
Figure 6. Mean spectra of maximum and minimum normalized force NK
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 857
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 7. Comparison between earthquake response of case models I and IV subjected to the Newhall ground motion (

"0)5 sec,
["7, :"0)2)
with peak errors of about 8 per cent. However, the inclusion of the vertical ground motion component and
the complete dynamics of the system in the exact model, leads to structural peak shear values signicantly
larger (17)4 per cent) than obtained from the simplied model. Such is also the case with the normal contact
force NK, which reaches a maximum 91 per cent larger than the corresponding unitary value predicted by the
simplied model. The eect of the vertical ground motion component is also apparent in the waviness of the
forcedeformation loop presented. As it was mentioned earlier, the spikes of these loops result from an
increase in the normal contact force due mainly to the vertical ground motion component; peak errors in the
FPS force range between 13 and 20 per cent.
An interesting aspect of Figure 7 is that the peak value of N, and hence of f (equation (2)), tend to occur at
large values of the lateral deformation of the FPS. This implies an increase in the maximum value attained by
the horizontal FPS restoring force f, which in turn leads to an increase in the maximum base shear of the
structure. This increase is due to the simultaneous occurrence of the maximum lateral deformation and
a peak positive vertical acceleration of the structure. It so happens that this simultaneous occurrence is
indirectly related to the statistical correlation between horizontal and vertical ground motion components.
Results from this study show that ground motion components with horizontal to vertical correlation larger
than about 0)15 will tend to present this eect.`
858 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 8. Peak deformation and base shear error spectra for the six ground motions considered using case model I (["7)
Mean and mean plus (minus) one standard deviation spectra of the errors in isolator deformations and
structure base shearof case model I relative to IVfor the six ground motions considered are presented in
Figure 8. Mean deformation errors are usually less than 5 per cent and, hence, are reasonably predicted by
the approximate model. Mean plus one standard deviation deformations are usually less than 15 per cent; the
variation increases with increasing period. The errors are usually positive indicating that the simplied model
predicts larger values than the exact model. On the other hand, mean values of base shear are under-
estimated by less than 10 per cent using the approximate model. Mean-plus-one standard deviation results
showunderestimations of base shear as large as 30 per cent. Indeed, these variations may be as large as 40 per
cent for specic ground motions where correlation exists between the lateral and vertical ground motion
components.
The accuracy of intermediate structural case models II and III presented earlier in Table I is evaluated
relative to the exact case model IV. Shown in Figure 9 are the error spectra corresponding to each of the four
models used to compute the isolator deformation and base shear in the structure subjected to three of the six
ground motions considered. As shown in the gure, the accuracy of the dierent building models is
dependent on the ground motion selected. For instance, errors in isolator deformations range in general from
0 to 20 per cent, and are less than 15 per cent for structures with xed-base period less than 1 sec. These
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 859
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 9. Peak deformation and base shear error spectra for the structural models considered in Table I subjected to Sylmar, Newhall
and Corralitos (["7, :"0)2)
prediction errors in deformation occur almost always in the safe side, i.e., simplied models tend to
overestimate the isolator deformation demand. On the other hand, the predicted base shear by simple models
can be either smaller or larger than the actual base shear depending on the model and ground motion
considered. For instance, case model II tends to overestimate the base shear in presence of large deformations
as produced by the Sylmar and Newhall ground motions. Prediction errors are as large as 30 per cent and
tend to be larger for structures with xed-base period less than 1 sec. Results are in general insensitive to the
vertical-to-lateral stiness ratio [,` for typical building values of this parameter (say greater than 5). This does
not imply, however, that the vertical dynamics of the FPS could be neglected; it rather implies that the most
relevant eects in the vertical motion of the system results from the consideration of the vertical ground
motion component and the lateralvertical coupling induced by the sliding on the spherical surface.
Another nal aspect considered is the eect of bidirectional input motion. Shown in Figure 10 is
a comparison between the predicted earthquake response of a structure with parameters

"0)3 sec,
vertical to lateral stiness ratio ["7, mass ratio :"0)2, and subjected to the Corralitos ground motion.
Results are presented for the exact structural model. The peak deformation of the isolation level when the
analysis of the structure is performed in the y-direction will be denoted as y
'
. Given the symmetry of the FPS
system, this radial deformation is presented as a circular boundary (Figure 10) in the deformation space
described by the horizontal deformation of the base. Superimposed to that boundary is the response of the
structural model subjected to both horizontal ground motion components. Thus, FPS deformations beyond
the circular boundary of radius y
'
represents an increase in the actual deformation due to the bidirectional
860 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 10. Comparison between the unidirectional and bidirectional earthquake response of the system (

"0)3 sec, ["7, :"0)2)

subjected to the Corralitos ground motion
input motion. Also presented in the gure is a comparison between the FPS deformation spectra for a single
(dashed-line) and bidirectional (solid line) ground motion components. The errors obtained by ignoring the
eect of bidirectional ground motion may be as large as 30 per cent in this case, and are specially large for sti
structures, which are precisely those most adequate for the use of an isolation system.
In order to generalize these results, error spectra were computed using the exact model for the six ground
motion records considered in this investigation. The mean and mean plus (minus) one standard deviation
error spectra are presented in Figure 11 for three dierent values of the mass ratio, :"0.1, 0.2, and 0)5, and
vertical to lateral stiness ratio ["7. It is apparent from the gure that errors of ignoring the two ground
motion components in the analysis are larger for rigid structures and decrease steadily with increasing
period. Mean errors in the estimated FPS deformations are about 20 per cent; mean plus (minus) one
standard deviation may be as large as about 38 per cent. These results seem to be rather insensitive to the
vertical-to-lateral stiness ratio [.`
A PHYSICAL MODEL OF THE FPS
So far, the FPS isolator has been assumed to stay at all times in contact with the concave sliding surface
during the earthquake. To include the possible uplift and impact between the structure and sliding surface, it
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 861
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 11. Error spectra of normalized FPS horizontal deformation (vectorial) for the six ground motions considered (["7, :"0)1,
0)2 and 0)5)
becomes useful to dene a physical model of the FPS isolator that simplies signicantly the analytical
formulation and solution of this problemotherwise such solution would be unnecessarily cumbersome.
The only addition of this model relative to the one presented in Figure 2 is that it considers a uniaxial gap
element between isolator and the sliding surface. By knowing the force in such element it is possible to keep
track of the uplift and impact of individual isolators.
The form in which isolator uplift is modeled deserves special attention. Most commonly, isolator uplift will
occur in fewisolators rather than all simultaneously. In that case, a formulation based on relative coordinates
to the ground is appropriate. Such formulation implies that the position of the sliders relative to the sliding
surfaces is known at all times, and so is the position where the slider comes in contact again with the sliding
surface after uplift. In the unusual case that all sliders undergo uplift, the formulation of the problem should
be based in absolute coordinates in order to known the exact position and velocity of the ground and
structure at the time of impact between the two. As conventionally done, a restitution coecient is
introduced to account for the energy loss during the impact in the corresponding isolators.
For the sake of brevity, the analytical formulation of the physical model will be omitted here since it will be
presented in a follow-up paper. However, it seems appropriate to comment on the advantages and
limitations of such model. The most obvious advantage of the physical model is that enable us to compute
the normal contact force N in a simple way. The force developed in the vertical gap element corresponds to
862 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
the vertical component of the total reaction R acting on the slider (Figure 1(c)). Since we know the direction
of R and one of its components (the vertical one), this reaction can be simply determined. Moreover, this
model is capable of simulating the uplift and impact of the slider through the gap element. It only requires
that the integration step be adapted dynamically at the instances of impact to attain sucient accuracy. The
main disadvantage of the model is the existence of a high vertical frequency associated with the large vertical
stiness of the isolator, which requires of an unconditionally stable integration procedure to avoid stability
problems in obtaining the solution. Another interesting advantage of this physical model, is that it allows to
consider the FPS as a typical non-linear element, in which the input is the deformation and velocity and the
output is the restoring force f. Thus, this physical model can be incorporated into a conventional structural
inelastic analysis program by considering both, the lateral degrees of freedom of the superstructure and base,
and the vertical degrees of freedom of each isolator.
In order to understand the actual performance of structures isolated with the FPS, i.e., including eects
such as large deformations, uplift, impact, and the true vertical dynamics of the system, the building example
presented in Figure 2(b) was studied in detail. The structure is a typical four-storey and two bay reinforced
concrete frame with xed-based fundamental period

"0)53 sec, constant modal damping ratio "0)05,

and j"0)07. Two structural cases are considered: (1) a system that includes the foundation beams presented
in Figure 2(b) (30 cm;70 cm) and having mass ratio :"0)25, and (2) a system without the foundation beams
and having :"0.` The structure is supported on three FPS isolators with radii R"100 cm, leading to
a nominal isolated period of 2 sec in both structural models; the resulting [ values for the two structures
considered are 10 and 13, respectively. As conventionally assumed, beams in the structure were considered
axially rigid, implying a single horizontal degree of freedom per building oor. Foundation beams, however,
were considered with their nominal axial as well as bending stiness in order to avoid unnecessary
simplications at the isolation interface. Furthermore, all vertical as well as rotational degrees of freedom
were considered in the structure and a consistent mass formulation was used in order to keep the dynamic
model as close to reality as possible. A seismic weight of 10 kN/m` was considered in every storey of the
model; also, a coecient of restitution c
'
"0)9 was considered in modeling the isolator impact. As before, two
structural models were analysed, one denoted as exact which includes all the above modelling assumptions
and one denoted as approximate which ignores the vertical dynamics as well as the overturning of the
building. These two cases are equivalent to case models I and IV presented earlier in Table I but extended to
the multistorey case.
An important aspect in the modelling of this system is the consideration of gravitational loads. First,
a static vertical load analysis of the structure was performed to compute all member forces including the ones
in the gap elements of the FPS isolators. The results of that analysis determined a pre-compression condition
for the gap elements, necessary in order to check the uplift of individual isolators. Of course, the acceleration
of gravity g needs to be included in the vertical dynamics of the problem; indeed, this acceleration is
responsible for bringing back to the ground the isolator and structure after been uplifted.
The dynamic response of the system subjected to the Newhall record is presented in Figures 12 and 13 for
both values of mass ratio :"0)25 and 0. Although results are presented here only for the Newhall ground
motion record, the trends observed in this case carry on to other near-eld input motions with a strong initial
acceleration pulse, such as Sylmar; they are also observed but in a lesser degree for higher-frequency motions
such as Llolleo and Vin` a. There are a number of interesting and practical issues in these gures that should
be accounted for in the design of structures supported by FPS isolators.
First, it is observed that discrepancies between the x-direction isolator deformations of the two structural
models are small, less than 6 per cent, which is consistent with the results presented earlier in Figure 9. Such
discrepancy increases to about 32 per cent when considering the base shear in the structure, which is also
consistent with the trends observed for the single FPS system (Figures 8 and 9). Second, it is interesting to
observe that uplift occurs in isolators under columns 1 and 3, as it is shown by the zero normal force attained
in these isolators at dierent instants of time. Uplift is identied at several instants of the response in spite of
the rather large seismic mass considered and the small height-to-base aspect ratio of the structure. Notice
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 863
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 12. Earthquake response of four-storey frame subjected to the Newhall ground motion (

"0)55 sec, ["10, :"0)25, and

c
P
"0)9). (a) global response; (b) column and FPS forces
864 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
Figure 13. Earthquake response of four-storey frame subjected to the Newhall ground motion (

"0)71 sec, ["13, :"0, and

c
P
"0)9). (a) global response; (b) column and FPS-forces
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 865
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
that uplift and impact occur simultaneously with the largest isolator deformation (see the bottom force
deformation loops). As a result of the uplift and downward impact, the maximum normal contact force
N ranges from 2 to 3)7 times the actual gravitational load in the isolators of the structure with :"0)25 and
from 2)2 to 5)8 times the dead load in the isolators of the structure with :"0. More important, the shear in
the building columns is increased signicantly due to the increase in the normal contact force N and, hence,
in the FPS horizontal restoring force f. Columns shears range from 0)37 to 0)56 times the dead load in the
columns in the rst system, leading to an increase of over 50 per cent relative to the estimated shears using the
simplied model. Such eect is dramatically increased in the structure with :"0 as a result of not having the
ltering action caused by the base mass and the lateral restraint of the foundation beams responsible for the
uniform distribution of the base shear among columns. The values of column shear in the latter case range
from 0)42 to 1)08 times the dead load in the columns, which represents a peak increase of about 300 per cent
over the column shear values obtained from the simplied model. Such large column base shears result from
the instantaneous sticking produced in the isolator due to the increase in the normal contact force resulting
from the downward impact and overturning of the structure.
Third, it is interesting to note by comparing Figures 12 and 13 that although global building responses,
such as base deformations and building base shear are similar in both structures analyzed, element
responses, such as individual isolator deformations are substantially dierent between the two structures
considered. The already mentioned eect of instantaneous isolator sticking, due to the increase in normal
force as a result of impact and overturning, implies a xed-base condition for the corresponding column at
that instant. If this occurs while the structure without the foundation beams is displacing in one direction, the
drift in the column that sticks may be increased considerably. Indeed, such column drift is almost 3 times
larger in the latter.
CONCLUSIONS
This investigation about modelling aspects of the FPS system has led to the following conclusions:
The predicted global response quantities, such as isolator and building deformations, obtained from the
dierent structural models considered in this investigation, are in general very similar. For instance, the
simpler methods overestimate these deformations in the mean by about 5 per cent. This conclusion agrees
with the results of previous analytical and experimental investigations.` " `
Larger discrepancies among models are observed, however, in building base shear and drift. In estimating
such response quantities, the rst key decision that needs to be made by the engineer involves estimating the
degree of correlation between the horizontal and vertical ground motion components for the given site of the
structure. If such correlation is expected to be small in absolute value, say less than 0)15, a simpler structural
model such as model 1 in Table I can still be used within 15 per cent error. However, the errors in the
estimated building base shear using model 1 and for structures subjected to correlated ground motion
components, may be as large as 30 per cent. Therefore, case models 24, which include the eect of the
vertical ground motion component, need to be considered in this case.
Moreover, if isolator deformations larger than about 15 cm are expected, models 3 and 4 (Table I)
including a vertical ground motion component and large deformations, should be used for dynamic analysis
of the structure. Although case model 3 is approximate, it leads to results that lie within 10 per cent error
relative to the exact solution obtained from model 4.
In such case also, estimations of member forces and deformations, should always be obtained from models
3 and 4, or alternatively, from a physical model like the one presented herein.
Normal contact forces during an earthquake may vary in the mean from one-fth to twice the axial load
due to gravitational loads. This increase may be as large as 5 if there is uplift in the isolators resulting from
overturning of the structure and vertical ground acceleration. The resulting vertical impact of the slider and
the spherical surface leads to two eects that need to be evaluated carefully.
866 J. L. ALMAZA N, J. C. DE LA LLERA AND J. A. INAUDI
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)
First, column base shears may increase due to the increase in normal force at the isolator interface due to
overturning and vertical impact. The resulting instantaneous increase in normal force makes the slider to
stop from sliding and to transmit signicantly larger shear forces to the supported columns. The eect
becomes more important when the mass of the isolation system is small and there are no foundation beams
(or a slab) connecting the isolators. It then seems advisable that independently isolated columns be braced
and interconnected by a signicant mass at the isolation interface capable of ltering the large increase in
normal force due to the potential vertical impact resulting from uplift.
Second, variations in the normal contact force lead also to non-uniform drift demand among unbraced
building columns due to the sudden restriction to sliding encountered by the columns with increased N.
Consequently, uplift should be impaired in order to avoid such localized increase in drift demand.
The Frictional Pendulum System is a reliable device to reduce the earthquake demand on a structure.
However, although local eects such as the variation in normal contact forces, large deformations, and uplift
do not seem to aect considerably the global system response, they must be considered in the isolation
modelling and design, especially for near-eld earthquakes with a strong initial acceleration pulse and for
statistically correlated horizontal and vertical ground motion components.
ACKNOWLEDGMENTS
This investigation has been funded by the Chilean National Research Fund for Science and Technology,
FONDECYT through Grant No. 1950987 and No. 1971078, and by the National Fund for the Foment of
Science and Technology, FONDEF through Grant No. D96I1008. The authors are grateful for this support.
REFERENCES
1. M. Constantinou, A. Mokha and A. Reinhorn, Teon Bearings in Base Isolation, Part II: Modeling, J. Struct. Engng. ASCE 116,
455474 (1990).
2. A. Mokha, N. Amin, M. Constantinou and V. Zayas, Seismic isolation retrot of large historic buildings, J. Struct. Engng ASCE 122,
298308 (1996).
3. V. Zayas, S. Low and S. Mahin, The FPS earthquake resisting system, Report CB/EERC-87/01, Earthquake Engineering
Research Center, University of California at Berkeley.
4. V. Zayas, S. Low, L. Bozzo and S. Mahin, Feasibility and performance studies on improving the earthquakes resistance of new and
existing buildings using the frictional pendulum system, Report CB/EERC-89/09, Earthquake Engineering Research Center,
University of California at Berkeley.
5. V. Zayas, S. Low and S. Mahin, A simple pendulum technique for achieving seismic isolation, Earthquake Spectra 6, 317334 (1990).
6. A. Mokha, M. Constantinou and A. Reinhorn, Experimental study of friction-pendulum isolation system, J. Struct. Engng. ASCE
117, 12011217 (1991).
7. T. Al-Hussaini, V. Zayas and M. Constantinou, Seismic isolation of multi-storey frame structures using spherical sliding isolation
system, Report NCEER-94-0007, National Center for Earthquake Engineering Research, State University of New York at Bualo.
8. J. Almaza n and J. De la Llera, Lateral and torsional earthquake behavior of structures isolated with the frictional pendulum system,
Report DIE niversidad Cato lica de Chile, in preparation.
STRUCTURES ISOLATED WITH THE FRICTIONAL PENDULUM SYSTEM 867
1998 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 27, 845867 (1998)